Storage
On This Page
Overview
Betty has many filesystems working together to provide a cohesive storage experience for your research data needs. It is important to understand where your data is to understand what to expect. This tutorial provides a brief overview.
Home Directories
Each user gets a HOME directory setup on the VAST Filesystem this provides high performance and reliability for your most important files such as code repositories, scripts, and configuration files.
| Location | /vast/home/<first letter of PennKey>/<PennKey> |
| Quota Size | 50GB |
| Quota INodes | 250,000 |
Users must be careful, if the HOME directory is filled users may not be able to connect. HOME directories have bi-weekly snapshots, but we highly recommend keeping a copy of these files elsewhere such as GitHub, Google Drives, or your laptop.
VAST Projects
VAST is a high performance flash file system. Users are strongly encouraged to read/write to this file system if your application is heavily dependent on I/O of you users expect high performance processing (large amounts of data). By default all PIs will have an allocation of 50GB to provide their group a small amount of community working space.
| Location | /vast/projects/<PI-PennKey>/default |
| Quota Size | 50GB |
| Quota INodes | 250,000 |
Additional quotas can be requested for additional projects.
| Location | /vast/projects/<PI-PennKey>/<project name> |
| Quota Size | # GB |
| Quota INodes | # GB * 5000 |
Ceph Projects
Ceph is a distributed filesystem configured with HDDs and a small front end NVMe cache. This is designed for “long term storage file system”. Users should place files that are not active (accessed by computing jobs) on this file system.
| Location | /ceph/projects/<PI-PennKey>/<project name> |
| Quota Size | # GB |
| Quota INodes | # GB * 500 |
Local Scratch
Compute nodes have a modest amount of local scratch that can be used if fast I/O is needed. It is located in /tmp
| Location | /tmp/$USER |
| Shared Size | ~500GB |
From a compute node, you can make a folder in the local scratch space.
mkdir /tmp/$USER
Once your compute is done ensure you copy the data and remove from the local scratch.
cp -r /tmp/$USER <project or home folder>
rm -r /tmp/$USER